\begin{tabular}{l c c }
\hline \hline
 & (1) & (2)  \\ 
\cline{2-3}
 \hspace{10pt} Math over-confidence &  0.118$^{***}$ & 0.116$^{***}$   \\
 \hspace{10pt}   & (0.031) & (0.032)  \\
\hspace{20pt} N & 1747 &    \\ 
\hspace{20pt} Sample mean & 0.041 &    \\ \\
 \hspace{10pt} Math under-confidence &  0.042$^{*}$ & 0.042$^{*}$   \\
 \hspace{10pt}   & (0.024) & (0.025)  \\
\hspace{20pt} N & 1747 &    \\ 
\hspace{20pt} Sample mean & 0.063 &    \\ \\
 \hspace{10pt} Math confidence (standard deviation units) &  0.182$^{***}$ & 0.185$^{***}$   \\
 \hspace{10pt}   & (0.025) & (0.025)  \\
\hspace{20pt} N & 1747 &    \\ 
\hspace{20pt} Sample mean & 0.000 &    \\ \\
Basic controls:                             & \checkmark  & \checkmark           \\
Added background controls:      &                    & \checkmark    \\
\hline \hline 
\end{tabular}
\begin{tabular}{ p{6in} }
\footnotesize
Notes:  This table regresses adolescent confidence outcomes on various definitions of childhood math confidence with various                  controls. Adolescent confidence is measured five years after the childhood measurement. In each row, the dependent variable is the adolescent measurement of the independent variable described. The measures of over- and under-confidence are our main binary measures. Our secondary measure of degrees of confidence takes on values from -6 to 6 and persistence of that variable is shown in the third row. The fourth row standardizes the degrees of confidence measure to have mean 0 and standard deviation 1, to facilitate ease of interpretation.                  All controls that are time-variant are observed in the same year as the confidence measures.                  Basic controls  include child gender, race, decile fixed effects for math and reading test percentile                  scores, digit span test scores, a general confidence index, family taxable income and its square, parent education,                  quarter-of-birth fixed effects, year-of-birth fixed effects, age at which confidence was measured fixed effects,                  and state fixed effects. We also include fixed effects for adolescent test score deciles in math and reading.   Added background controls are parents' rating of child health,                  indicators for receiving government transfers, household structure, parenting practices, parent occupation,                 and parent mental health and confidence measures. All controls are recoded to zero if missing and we include a missing indicator.           Standard errors are clustered by family, and included in parentheses below each          estimate. *, **, and *** indicate significance at the 0.1, 0.05, and 0.01 percent level, respectively. 
\end{tabular}
